A robust iterative algorithm for image restoration

We present a robust iterative method for the restoration of images that are degraded by blur and noise. The key to our approach is a new iterative energy function that can regularize the gradient in the steepest descent by adaptively adjusting to the current state in the iterative process, so that the iterative solutions converge to the real scene despite noise contamination in the observation. This approach is general and can be implemented by combining the VanCittert iterative algorithm with a well established noise reduction method. Numerical experiments show our method has a good balance between structure recovery and noise reduction and performs favorably in two tests compared to three existing regularization methods.